Dash

Data agent that grounds SQL analysis in five context layers and can save fixes as learnings.

Dash is a data agent that grounds SQL analysis in five context layers and can save fixes as learnings.

Ask a question in English and get a correct, meaningful answer. That’s the goal. But raw LLMs writing SQL hit a wall fast: schemas lack meaning, types are misleading, tribal knowledge is missing, there’s no way to learn from mistakes, and results lack interpretation. The root cause is missing context and missing memory. Dash solves this with five layers of grounded context, a learning loop that can save fixes for later queries, and a focus on delivering insights you can act on.

Chat with Dash in Slack, the terminal, or the AgentOS UI. The code is public at agno-agi/dash.

How it works

Dash runs as an Agno team in coordinate mode, with a leader that routes each request to two specialists:

Agent	Role
Analyst	Queries company data; use a SELECT-only role for enforced read-only access
Engineer	Builds reusable views and summary tables in the dash schema
Leader	Routes queries, coordinates the team, posts to Slack

Schema boundaries: Company data lives in the public schema; agent-created views and summary tables live in the dash schema. The Analyst connection uses default_transaction_read_only=on, so each transaction starts read-only. A writable role can switch a transaction to read-write, and the bundled ai role is writable. Use a separate SELECT-only role to enforce the Analyst boundary. The Engineer uses the application’s database role, and a regex-based SQLAlchemy listener blocks common writes to public. That listener is a best-effort guard, not a database privilege boundary. Use a role without public write privileges when the boundary must hold for every SQL form.

Five layers of context

Layer	Purpose	Source
Table Usage	Schema, columns, relationships	knowledge/tables/*.json
Human Annotations	Metrics, definitions, business rules	knowledge/business/*.json
Query Patterns	SQL that is known to work	knowledge/queries/*.sql
Learnings	Error patterns and discovered fixes	Agno Learning Machine
Runtime Context	Live schema changes	introspect_schema tool

Self-learning

The Analyst is instructed to retrieve knowledge and learnings before generating SQL. It can diagnose a failed query and save the fix as a learning. These tools are model-callable, so retrieval and saving depend on the model’s tool choices.

System	Stores	How it evolves
Knowledge	Validated queries, table schemas, business rules	Curated by you and refined by Dash
Learnings	Error patterns and discovered fixes	Added when Dash calls save_learning

When a churn query uses the wrong filter, Dash can save the corrected ended_at IS NULL pattern for later retrieval. When your team defines MRR as the sum of active subscriptions excluding trials, that rule can live in knowledge/business/ for later queries.

Insights you can act on

Dash reasons about what makes an answer useful. Ask “Which plan has the highest churn rate?” and you get the number, the comparison across plans, the trend behind it, and any caveats from your business rules.

Run locally

git clone https://github.com/agno-agi/dash.git && cd dash

cp example.env .env
# Edit .env and add your OPENAI_API_KEY

docker compose up -d --build

# Generate sample data and load knowledge
docker exec -it dash-api python scripts/generate_data.py
docker exec -it dash-api python scripts/load_knowledge.py

Confirm Dash is running at http://localhost:8000/docs. The Dash README walks through this step by step.

Connect to the AgentOS UI

1. Open os.agno.com and log in.
2. Click Connect OS, choose Local, and enter http://localhost:8000.
3. Click Connect.

Dash is running locally.

Deploy to Railway

Railway deployment uses .env.production to keep production credentials separate from local dev.

cp example.env .env.production
# Edit .env.production and set OPENAI_API_KEY

1. Deploy infrastructure

   railway login
   ./scripts/railway_up.sh
   

   This creates the Railway project, database, and app service. The app will crash-loop until the JWT key is added in the next step. That’s expected.

2. Get your JWT key

   Production requires a JWT_VERIFICATION_KEY from AgentOS. You need the Railway domain from step 1 to set this up.

   1. Copy your Railway domain from the output of step 1 (e.g. dash-production-xxxx.up.railway.app).
   2. Open os.agno.com and log in.
   3. Click Connect OS, choose Live, and paste your Railway URL.
   4. Go to Settings → OS & Security and turn on Token-Based Authorization (JWT). The UI generates a key pair and shows you the public key.
   5. Leave JWT_VERIFICATION_KEY unset in .env.production. Set it directly on the dash service, preserving the PEM line breaks:

   railway variables --set "JWT_VERIFICATION_KEY=-----BEGIN PUBLIC KEY-----
   MIIBIjANBgkq...
   -----END PUBLIC KEY-----" --service dash
   

   The current Dash scripts cannot safely share a multiline PEM through .env.production. railway_up.sh sources the file and requires shell quoting, while railway_env.sh leaves a trailing quote in a quoted multiline value.

3. Push environment and redeploy

   The current railway_env.sh exits after its first variable because set -e treats the initial ((count++)) result as a failure. Replace ((count++)) with ((++count)) in scripts/railway_env.sh before running it.

   ./scripts/railway_env.sh
   ./scripts/railway_redeploy.sh
   

   After this change, railway_env.sh reads the remaining values from .env.production and sets them on the Railway service. It is safe to run repeatedly.

Dash is live on Railway.

The Dash README covers this flow in more detail.

Production operations

Database scripts must run inside Railway’s network. The internal hostname pgvector.railway.internal is unreachable from your local machine, so SSH into the running container:

railway ssh --service dash
# Inside the container:
python scripts/generate_data.py
python scripts/load_knowledge.py

Other operations run locally:

railway logs --service dash
railway open

Connect to Slack

Dash can receive DMs, @mentions, and thread replies, and can post to channels proactively. Each Slack thread maps to one Dash session.

1. Run Dash with a public URL (ngrok locally, or your Railway domain).
2. Create and install the Slack app from the manifest in docs/SLACK_CONNECT.md.
3. Set SLACK_TOKEN and SLACK_SIGNING_SECRET, then restart Dash.
4. In Slack, confirm Event Subscriptions shows verified, then send a DM or @mention to test.

See the Slack setup guide for the manifest, ngrok commands, permissions, and troubleshooting.

Example prompts

Try these on the sample SaaS metrics dataset:

• What’s our current MRR?
• Which plan has the highest churn rate?
• Show me revenue trends by plan over the last 6 months
• Which customers are at risk of churning?

Add your own data

Dash works best when it understands how your organization talks about data:

Directory	Content
knowledge/tables/	Table meaning, column notes, data quality caveats
knowledge/queries/	Proven SQL patterns
knowledge/business/	Metric definitions, business rules, common gotchas

Load or update knowledge at any time:

python scripts/load_knowledge.py             # Upsert changes
python scripts/load_knowledge.py --recreate  # Fresh start

The Dash README covers loading your own data and scheduled proactive tasks.

Run evals

Five eval categories using Agno’s eval framework:

Category	Eval type	What it tests
accuracy	AccuracyEval (1-10)	Correct data and meaningful insights
routing	ReliabilityEval	Team routes to the correct agent and tools
security	AgentAsJudgeEval (binary)	No credential or secret leaks
governance	AgentAsJudgeEval (binary)	Refuses destructive SQL operations
boundaries	AgentAsJudgeEval (binary)	Schema access boundaries respected

python -m evals                      # Run all evals
python -m evals --category accuracy  # Run specific category
python -m evals --verbose            # Show response details

Source

Dash is public at agno-agi/dash. The README covers the full architecture, the data model, and the security setup.